The Conduit Artery Functional Evaluation (CAFE) Study Principal Results The CAFE investigators Bryan Williams, Peter S. Lacy, Simon McG. Thom, Kennedy Cruickshank, Alice Stanton, David Collier, Alun D. Hughes, Herbert Thurston. Study Advisor; Michael O’Rourke For the ASCOT Investigators

CAFE Study Steering Committee Professor Bryan Williams, MD, FRCP, FAHA Principal Investigator and Chairman of the Steering Committee University Department of Cardiovascular Sciences University of Leicester, Leicester, UK Dr Peter S. Lacy, PhD Steering Committee and CAFE Study Coordinator Professor Kennedy Cruickshank Clinical Epidemiology Group, University Department of Medicine, Manchester Royal Infirmary Manchester, UK Professor Alice Stanton Department of Clinical Pharmacology, Royal College of Surgeons in Ireland, St Stephen’s Green, Dublin, Ireland Dr David Collier Department of Clinical Pharmacology, William Harvey Research Institute, Bart’s & The London, Queen Mary’s School of Medicine & Dentistry, London, UK Professor Simon McG Thom Clinical Pharmacology and Therapeutics St. Mary’s Hospital Imperial College London London, UK Professor Herbert Thurston University Department of Cardiovascular Sciences University of Leicester, Leicester, UK ASCOT Sub-Study Committee Liaison: Prof. Eoin O’Brien, Prof. Mark Caulfield Study Advisor Professor Michael O’Rourke University of New South Wales, Sydney, Australia Statistical Support, CSRI, Gothenburg, Sweden B Dahlof, H Wedel, R Chamberlain, NG Pehrsson, M Molin, A Pivodic, J Lindqvist, A Hagelin For the CAFE Investigators and ASCOT Investigators

Background (1) Brachial blood pressure is a strong predictor of clinical outcomes in people with hypertension It is assumed that brachial blood pressure accurately reflects pressures in the central aorta and thus left ventricular load This assumption may not be valid in all circumstances because different classes of blood pressure–lowering drugs may differentially influence central aortic blood pressures

Background (2) In clinical trials comparing different blood pressure lowering–drugs, clinical outcomes could be influenced by drug effects on central aortic pressures, despite similar effects on brachial blood pressure This hypothesis required testing in a prospective clinical outcomes trial evaluating 2 different blood pressure– lowering treatment regimens The Conduit Artery Functional Evaluation (CAFE) study was designed to test this hypothesis as a major substudy of the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT)

CAFE Hypothesis Primary Objective The different blood pressure–lowering regimens in ASCOT (atenolol ± thiazide versus amlodipine ± perindopril) would produce different effects on central aortic pressures and haemodynamics despite similar effects on brachial blood pressure Secondary Objective Central aortic pressures would be an important determinant of clinical outcomes Williams B. and O’Rourke M. J Hum Hypertens. 2001;15(suppl 1):S69-S73.

CAFE Study Design Recruitment from 5 large UK and Ireland ASCOT study centres Central co-ordinating centre (Leicester, UK) Recruitment into CAFE commenced when patients were stable after up-titration of ASCOT medication (~1 year after start of ASCOT) 80% patients had more than one tonometry measurement Average number of tonometry measurements/patient was 3.4 Average follow-up after 1st measurement was 3.0yrs Circulation. 2006;113:1213-1225

Brachial Blood Pressure Pulse Wave Analysis Brachial Blood Pressure Bone Artery Sensor 140 70 Radial Transfer function Central Aortic 140 70

CAFE Study Profile 2199 subjects recruited from 5 UK ASCOT centres 126 excluded due to heart rate irregularity/poor waveforms 2073 evaluable for tonometry 1042 received amlodipine-based regimen 1031 received atenolol-based regime 4 subjects incomplete information, 1 alive at last visit, 2 withdrawn consent, 1 lost to follow-up 1 subject incomplete information, withdrawn consent 1042 assessed on an intention-to-treat basis 1038 complete information (997 alive, 41 dead) 1031 assessed on an intention-to treat basis 1030 complete information (989 alive, 41 dead) Circulation. 2006;113:1213-1225

Baseline Demographics (1) CAFE ASCOT Atenolol-based n=1031 Amlodipine-based n=1042 Atenolol-based n=9639 Amlodipine-based n=9618 Women 189 (18.3%) 208 (20%) 2257 (23.5%) 2258 (23.4%) Age (years) 62.6 (8.3) 62.9 (8.2) 63.0 (8.5) White 886 (85.9%) 892 (85.6%) 9170 (95.3%) 9187 (95.3%) Current smoker 251 (24.3%) 267 (25.6%) 3109 (32.3%) 3168 (32.9%) SBP mm Hg 159.9 (16.6) 161 (18.4) 163.9 (18) 164.1 (18.1) DBP (mm Hg) 92.4 (9.6) 92.6 (9.8) 94.5 (10.4) 94.8 (10.4) Heart rate (bpm) 71.8 (12.3) 71.2 (12.4) 71.8 (12.6) 71.9 (12.7) BMI (kg/m2) 29 (4.5) 29.1 (4.7) 28.7 (4.5) 28.7 (4.6) Height (cm) 170.7 (8.7) 170.2 (9.4) NA Total cholesterol (mg/dL) 224.3 (38.7) 224.3 (42.5) 228.2 (42.5) LDL-cholesterol (mg/dL) 143.1 (34.8) 146.9 (38.7) HDL-cholesterol (mg/dL) 50.3 (15.5) Triglycerides (mg/dL) 159.4 (88.6) 159.4 (8.6) 168.3 (88.6) Glucose (mg/dL) 110 (38) 112 (38) Creatinine (mg/dL) 1.08 (0.18) 1.09 (0.19) 1.09 (0.19) 1.09 (0.18) Need to improve this table – not clear Circulation. 2006;113:1213-1225

Baseline Demographics (2) CAFE ASCOT Atenolol-based n=1031 Amlodipine-based n=1042 Atenolol-based n=9639 Amlodipine-based n=9618 Medical history Prior stroke/TIA 76 (7.4%) 101(9.7%) 1063 (11.1%) 1050 (10.9%) Diabetes 252 (24.4%) 251 (24.4%) 2578 (26.8%) 2567 (26.6%) LVH (echo or ECG) 237 (23%) 256 (24.6%) 2076 (21.6%) 2091 (21.7%) Peripheral vascular disease 61 (5.9%) 59 (5.7%) 613 (6.4%) 586 (6.1%) Other relevant CV disease 22 (2.1%) 27 (2.6%) 486 (5.1%) 533 (5.5) Mean (SD) # risk factors 3.7 (0.9) Drug therapy BP treatment naive 109 (10.6%) 100 (9.6%) 1825 (19%) 1841 (19.1%) Lipid-lowering therapy 120 (11.6%) 120 (11.5%) 1004 (10.4%) 1046 (10.9%) Aspirin use 244 (23.7%) 274 (26.3%) 1837 (19.1%) 1851 (19.2%) Need to improve this table Circulation. 2006;113:1213-1225

Calcium Channel Blocker Regimen -Blocker Based Regimen Treatment Algorithm Calcium Channel Blocker Regimen -Blocker Based Regimen Step 1 Amlodipine 5 mg Atenolol 50 mg Step 2 Amlodipine 10 mg Atenolol 100 mg Step 3 Amlodipine 10 mg + perindopril 4 mg Atenolol 100 mg + bendroflumethiazide K 1.25 mg Step 4 Amlodipine 10 mg + perindopril 8 mg (2x4 mg) Atenolol 100 mg + bendroflumethiazide K 2·5 mg Step 5 Amlodipine 10 mg + perindopril 8 mg (2x4 mg) + doxazosin gastrointestinal transport system 4 mg Atenolol 100 mg + bendroflumethiazide K 2·5 mg + doxazosin gastrointestinal transport system 4 mg Step 6 Amlodipine 10 mg + perindopril 8 mg (2x4 mg) + doxazosin gastrointestinal transport system 8 mg Atenolol 100 mg + bendroflumethiazide K 2·5 mg + doxazosin gastrointestinal transport system 8 mg Further treatment to achieve blood-pressure goal outlined at http://www.ascotstudy.org. All drugs given orally. Circulation. 2006;113:1213-1225

Difference (Atenolol-Amlodipine) Haemodynamic Data Parameter Atenolol Amlodipine Difference (Atenolol-Amlodipine) Statistics t-test (P) Brachial SBP (mm Hg) 133.9 (133, 134.7) 133.2 (132.5, 133.8) 0.7 (-0.4, 1.7) 0.2 Brachial DBP (mm Hg) 78.6 (78.1, 79.1) 76.9 (76.4, 77.4) 1.6 (0.9, 2.4) <.0001 Brachial PP (mm Hg) 55.3 (54.6, 56) 56.2 (55.6, 56.9) -0.9 (-1.9, 0) .06 Heart rate (BPM) 58.6 (58, 59.2) 69.3 (68.6, 69.9) -10.7 (-11.5, -9.8) Data: Area under Curve ± 95% CI Circulation. 2006;113:1213-1225

Brachial and Central Aortic Systolic Blood Pressure (± 95% CI) Brachial SBP Diff Mean (AUC) = 0.7 (-0.4,1.7) mm Hg 140 Atenolol Amlodipine 135 133.9 133.2 P=.07 130 mm Hg 125 125.5 121.2 120 P<.0001 Central SBP Diff Mean (AUC) = 4.3 (3.3, 5.4) mm Hg 115 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 AUC Time (Years) Atenolol 86 243 324 356 445 372 462 270 339 128 85 1031 Amlodipine 88 248 329 369 475 406 508 278 390 126 101 1042 Circulation. 2006;113:1213-1225

Brachial and Central Aortic Pulse Pressure by Treatment Arm Brachial PP Diff Mean (AUC) = -0.9 (-1.9, 0) mm Hg Atenolol Amlodipine 58 56.2 55.3 P=.06 53 mm Hg 45 46.4 43 43.4 Central PP Diff Mean (AUC) = 3 (2.1, 3.9) mm Hg P<.0001 38 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 AUC Time (Years) Atenolol 86 243 324 356 445 372 462 270 339 128 85 1031 Amlodipine 88 248 329 369 475 406 508 278 390 126 101 1042 Circulation. 2006;113:1213-1225

Augmentation Index (%) by Treatment Arm 40 Atenolol Amlodipine 35 31.9 30 Alx (%) 25 25.3 20 Diff Mean (AUC) = -6.5 (5.8, 7.3)% P<.0001 15 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 AUC Time (Years) Atenolol 86 243 324 356 445 372 462 270 339 128 85 1031 Amlodipine 88 248 329 369 475 406 508 278 390 126 101 1042 Circulation. 2006;113:1213-1225

Results Summary (1) Atenolol-based therapy was associated with higher central aortic systolic pressure and higher central aortic pulse pressure, despite similar brachial pressures, when compared with amlodipine-based therapy Central aortic outgoing pressure wave (P1 height) was lower with atenolol-based therapy vs amlodipine-based therapy Pulse wave augmentation and the percentage of the central aortic pressure wave attributable to wave reflection was increased by atenolol-based therapy compared with amlodipine-based therapy

Cox Regression Analysis Secondary Objective To define the relationship between central aortic pressures and haemodynamic parameters with clinical outcomes Composite Clinical Outcome Defined post-hoc when the ASCOT endpoint rate was known Included all cardiovascular events and procedures + development of renal impairment (ASCOT pre-specified and validated endpoints) 305 events All endpoints included from time of randomization into ASCOT Data analyzed blind to treatment allocation Cautious Interpretation Limited statistical power Circulation. 2006;113:1213-1225

Impact of Blood Pressure and Central Aortic Haemodynamics on Clinical Outcomes in the CAFE Study (Hazard/10 mm Hg) Updated Cox proportional hazard model for the composite endpoint, unadjusted Factor X2 P HR CI Peripheral PP 21.0 <.0001 1.21 1.12-1.30 Central PP 17.8 1.20 1.11-1.30 Augmentation 7.10 .008 1.22 1.06-1.40 P1 height 19.0 1.37 1.20-1.54 Updated Cox proportional hazard model for the composite endpoint, adjusted for baseline variables Factor X2 P HR CI Central PP 3.91 .048 1.11 1.00-1.21 Circulation. 2006;113:1213-1225

CAFE Study Conclusions (1) Despite similar brachial systolic blood pressure, amlodipine ± perindopril-based treatment was more effective than atenolol ± thiazide-based treatment at lowering central aortic systolic blood pressure and central aortic pulse pressure Brachial blood pressure is not always a perfect surrogate for the effects of drug therapy on central aortic pressures

CAFE Study Conclusions (2) Brachial blood pressure overestimated the haemodynamic benefit of atenolol ± thiazide-based treatment and underestimated the benefit of amlodipine ± perindopril-based treatment on central aortic pressures and haemodynamics Central aortic pressure may be an important independent determinant of clinical outcomes Results of the CAFE study suggest that the “central aortic blood pressure hypothesis” is a plausible mechanism to explain the better clinical outcomes for hypertensive patients treated with amlodipine ± perindopril-based therapy in ASCOT

The CAFE Study Investigators CAFE Investigators and CAFE Study Centers University Hospitals of Leicester Leicester Royal Infirmary, Leicester, UK B. Williams, P.S. Lacy, H. Thurston, D. O’Brien, P. Swales, S. Nanayakkara, A. Stanley, K. Edwards, W. Gamble, V. Garratt, J. Harris, A. Kirkham, Y. Revell-Smith, L. Bernhardt St. Mary’s Hospital, Paddington, London, UK S. McG Thom, A. Whitehouse, A. Hughes, J. Macduff, S. Leech, L. Kinsley, O. Trainor, Y. Chim, A. Jack, S. Murphy, R. Baruah, F. Harrison, R. Fernandez, A. Strain, M. Liboro The Beaumont Hospital, Dublin, Ireland A. Stanton, E. O’Brien, E. Dolan, D. Farrell, O.B. Gallagher, S. Lyons, G. McCarthy Manchester Royal Infirmary, Manchester, UK K. Cruickshank, S. Dean, J. Dunkerley, M. Banerjee, M. Holland, M. Hart, M. Cullen, L. Hardstaff, J. Reynolds, J. Collins, K. Peters, M. Luckson St. Bartholomew’s Hospital, London, UK D. Collier, M. Caulfield, C. Waleczko, G. Salahi-Ali, J. Sheil, B. Scott, S. Wilson, C. David, H. Fok